Blog

Field notes from building AI that has to work.

What we learned shipping real systems: the parts that broke, the numbers we measured, and what we'd do differently next time.

5 min read · August 14, 2026

Building Agents That Don't Break in Production

The demo took an afternoon. Making it survive real users took the other eleven weeks. What actually goes into an agent you can leave running.

Read article

4 min read · July 22, 2026

Your AI Feature Needs an Evaluation Set More Than a Better Prompt

Teams spend weeks tuning prompts and ship regressions anyway. The fix is boring and effective: measure quality as a number, in CI, before anything reaches a user.

4 min read · June 30, 2026

Most RAG Problems Are Retrieval Problems

When a grounded assistant gives a wrong answer, the model usually isn't at fault. It was handed the wrong context. Where to actually look.

1 min read · May 28, 2026

Product-Market Fit in the Era of AI

Generative AI is rewriting how teams find product-market fit. Here's what still holds true, and what's changing fast.

1 min read · May 20, 2026

Shipping Your First RAG Chatbot

Retrieval-augmented generation lets an assistant answer from your own documents. A practical look at what it takes to ship one.

1 min read · May 12, 2026

Why Your Startup Needs a Design System

A design system isn't a luxury for big companies. For a small team, it's the fastest way to ship a consistent, polished product.

1 min read · April 25, 2026

From Prototype to Production

The demo works on your laptop. Now comes the real work. What separates a convincing prototype from a product people can rely on.